摘要
为了获得更好的语音情感识别的实时性和正确率,该文提出了基于核典型相关分析和支持向量机的语音情感识别模型。首先提取多种情感识别的特征,采用核典型相关分析对特征进行选择,将选择的特征作为支持向量机的输入向量进行训练,建立情感识别的分类器,最后采用语音情感识别的标准数据库进行验证性和对比实验。实验结果表明,该模型能够准确识别不同类型的语音情感,获得较高的语音情感识别率。核典型相关分析减少了分类器的输入向量数,加快了情感识别速度,获得了理想的实时性。该文语音情感识别结果优于对比模型,具有更高的实际应用价值。
In order to obtain the better real-time and correct rate of the speech emotion recognition, an emotion recognition model based on the kernel canonical correlation analysis and the support vector machine is proposed here. Firstly, multiple features of the speech emotion recognition are extracted and the feature selection is selected by the kernel canonical correlation analysis, and then the selected features are taken as the input vector of the support vector machine to be trained for es-tablishing the classifier of the speech emotion recognition. Finally, experiments on the standard database of the speech emotion recognition is used to validate the performance of the model. The ex-perimental results show that, by using the kernel canonical correlation analysis with the less input vectors, the proposed model can accurately identify the emotion type and increase the recognition rate of the speech emotion,and has the better read-time. The result of the speech emotion recognition is better than that of the contrast models,and the model has the higher practical application value.
作者
张前进
王华东
Zhang Qianjin Wang Huadong(Department of Information Engineering,Anhui Vocational College of Defense Technology ,Lu＇an 237011,China School of Computer Science and Technology,Zhoukou Normal University,Zhoukou 466001,China)
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2017年第2期191-197,共7页
Journal of Nanjing University of Science and Technology
基金
2016年安徽省高等学校自然科学研究重点项目(KJ2016A120)
关键词
语音情感识别
核典型相关分析
特征选择
情感分类器
支持向量机
speech emotion recognition
kernel canonical correlation analysis
feature selection
emotion classifiers
support vector machine
作者简介
张前进（1982-）,男,硕士,讲师,主要研究方向：人工智能、数据挖掘,E-mail：zh-qj@foxmail.com
通讯作者：王华东（1977-）,男,硕士,副教授,主要研究方向：人工智能、数据挖掘,E-mail：46935563@qq.com.
